信号与系统实验

Signals and Systems Laboratory

Lab #2. Linear Time-Invariant
Systems
Instructor：张利君
Office: 一教129
Email: zhanglj@sustech.edu.cn
Department of Electrical and
Electronic Engineering
 Outline
1. Convolution Sum for Discrete-Time LTI
Systems
2. Causal LTI Systems Described by Linear
Constant-Coefficient Difference Equations
3. Convolution Integral for Continuous-Time LTI
Systems
4. Causal LTI Systems Described by Linear
Constant-Coefficient Differential Equations
5. Design a Wireless Signal Detector

信号与系统实验 Signals and Systems Laboratory 2

Part 1. Convolution Sum for
Discrete-Time LTI Systems

信号与系统实验 Signals and Systems Laboratory 3

 Representation of Discrete-Time Signals
+∞

𝑥𝑛 = 𝑥 𝑘 𝛿[𝑛 − 𝑘]
𝑘=−∞
n = 1:4;
x = [1 2 0.5 1];
stem(n,x,'filled'); xlabel('n');
title('x[n]');
xticks([1 2 3 4 ]); yticks([0 0.5 1
1.5 2]);
box off; grid on;

信号与系统实验 Signals and Systems Laboratory 4

 Representation of Discrete-Time Signals

+∞

𝑥𝑛 = 𝑥 𝑘 𝛿[𝑛 − 𝑘]
𝑘=−∞
Another example
Time index for
nonzero values

n = 11:14;
x = [1 2 0.5 1];
stem(n,x,'filled'); xlabel('n');
title('x[n]');
xticks([11 12 13 14]); yticks([0
0.5 1 1.5 2]);
box off; grid on;

信号与系统实验 Signals and Systems Laboratory 5

 Discrete-Time Unit Impulse Response

𝛿[𝑛 − 𝑘] S ℎ𝑘 [𝑛] (𝑘 = ⋯ , −1,0,1, ⋯)

S is linear
+∞ +∞
𝑥𝑛 = 𝑥 𝑘 𝛿[𝑛 − 𝑘] S 𝑦𝑛 = 𝑥[𝑘] ℎ𝑘 [𝑛]
𝑘=−∞ 𝑘=−∞
S is time-invariant
ℎ𝑘 𝑛 = ℎ[𝑛 − 𝑘]
+∞ +∞

𝑥𝑛 = 𝑥 𝑘 𝛿[𝑛 − 𝑘] S 𝑦𝑛 = 𝑥 𝑘 ℎ[𝑛 − 𝑘]
𝑘=−∞ 𝑘=−∞

Convolution Sum!

信号与系统实验 Signals and Systems Laboratory 6

Response of LTI Systems: Convolution Sum
+∞

𝑦𝑛 = 𝑥 𝑘 ℎ[𝑛 − 𝑘]
𝑘=−∞

𝑦 𝑛 = 𝑥[𝑛] ∗ ℎ[𝑛]

1. An LTI system is completely characterized by

its impulse response h[n].
2. From the impulse response, we can know
everything about the LTI system. (causality,
stability, static/dynamic, invertibility)
3. For any input, we can find the output
immediately by the convolution sum.

信号与系统实验 Signals and Systems Laboratory 7

 Calculate Convolution Sum
Example#1 𝑦 𝑛 = 𝑥[𝑛] ∗ ℎ[𝑛]

h 𝑛 = 𝛿 𝑛 + 𝛿 𝑛 − 1 + 𝛿[𝑛 − 2]
𝑥 𝑛 = 0.5𝛿 𝑛 + 2𝛿[𝑛 − 1]

① Inspect the time index k where 𝑥[𝑘] ≠ 0;

② Shift h[n] to ℎ[𝑛 − 𝑘] as a whole on the n
axis;
③ Multiply ℎ[𝑛 − 𝑘] by 𝑥 𝑘 ;
④ Sum over −∞ < 𝑘 < +∞, where 𝑥[𝑘] ≠ 0,
then we obtain the output 𝑦[𝑛] for all n as
a whole.

𝑦 𝑛 = 0.5ℎ 𝑛 + 2ℎ[𝑛 − 1]
This is convenient only for sequences
with short durations!

信号与系统实验 Signals and Systems Laboratory 8

 Calculate Convolution Sum
Example#2 𝑦 𝑛 = 𝑥[𝑛] ∗ ℎ[𝑛]
h𝑛 =𝑢 𝑛 𝑥 𝑛 = 𝛼𝑛𝑢 𝑛

① x k as a whole;
② Flip ℎ[𝑘] to ℎ −𝑘 ;
③ Shift ℎ[−𝑘] to ℎ[𝑛 − 𝑘] for each n;
④ Multiply 𝑥[𝑘] with ℎ[𝑛 − 𝑘] over the whole k axis;
𝑘
𝑥 𝑘 ℎ 𝑛−𝑘 = 𝛼 0≤𝑘≤𝑛
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
⑤ Sum over k, obtain the value of 𝑦[𝑛] for the specific time index n;
⑥ For other values of n, return to step 3 to do shift-multiplication-
summation again.
This is the usual way to
calculate convolution!

信号与系统实验 Signals and Systems Laboratory 9

 Convolution Sum in MATLAB: conv(u,v)
+∞
𝑛−𝑏 =𝑎 →𝑛 =𝑎+𝑏
𝑦 𝑛 =𝑢 𝑛 ∗𝑣 𝑛 = 𝑢 𝑘 𝑣[𝑛 − 𝑘] 𝑛−𝑏−𝑀 =𝑎+𝑁 →𝑛 =𝑎+𝑏+𝑀+𝑁
𝑘=−∞ ∴ 𝑛𝑦 = [𝑎 + 𝑏: 𝑎 + 𝑏 + 𝑀 + 𝑁]
 Time index for nonzero values of 𝑢 𝑛 : 𝑛𝑢 = 𝑎: 𝑎 + 𝑁 ▲= 𝑎 + 𝑏 = 𝑛𝑢 1 + 𝑛𝑣(1)
 Time index for nonzero values of 𝑣 𝑛 : 𝑛𝑣 = 𝑏: 𝑏 + 𝑀 ∎ = 𝑎 + 𝑏 + 𝑀 + 𝑁 = 𝑛𝑢 𝑒𝑛𝑑 + 𝑛𝑣(𝑒𝑛𝑑)
Total number of nonzero values of y[n] is:
 Time index for nonzero values of 𝑦 𝑛 : 𝑛𝑦 =▲: ∎ 𝑀+𝑁+1

𝑢[𝑘] ⋯ 𝑘 𝑣[𝑘]
𝑎
⋯ 𝑘
𝑎+1 𝑎+2 𝑎+𝑁 𝑏 𝑏+1 𝑏+𝑀

Flip w.r.t the origin

Shift to the right by n

𝑣[−𝑘] ⋯ 𝑘
−𝑏 − 𝑀 −𝑏 − 1 −𝑏

信号与系统实验 Signals and Systems Laboratory 10

 Convolution Sum in MATLAB: conv(u,v)
+∞

𝑦 𝑛 =𝑢 𝑛 ∗𝑣 𝑛 = 𝑢 𝑘 𝑣[𝑛 − 𝑘]
𝑘=−∞
 Define 𝑢, 𝑛𝑢 and 𝑣, 𝑛𝑣;
 Calculate 𝑦[𝑛] by 𝑦 = 𝑐𝑜𝑛𝑣(𝑢, 𝑣);
 Calculate the time index for nonzero values of 𝑦[𝑛]
𝑛𝑦 = 𝑛𝑢 1 + 𝑛𝑣 1 : 𝑛𝑢 𝑒𝑛𝑑 + 𝑛𝑣(𝑒𝑛𝑑)
u = [1 1 0.5 1.5];
nu = 3:6;
v = [1 1];
nv = 5:6;
y = conv(u,v);
ny = nu(1)+nv(1):nu(end)+nv(end);
subplot(3,1,1);
stem(nu,u,'r');xlim([3,12]);xlabel('n');
title('u[n]');
subplot(3,1,2);
stem(nv,v,'g');xlim([3,12]);xlabel('n');
title('v[n]');
subplot(3,1,3);stem(ny,y,'b');xlim([3,12
]);xlabel('n');title('y[n]=u[n]*v[n]');

信号与系统实验 Signals and Systems Laboratory 11

 Exercise 2.1

Calculate the convolution between the discrete-time signals

ℎ 𝑛 = 0.7𝑛 , 0 ≤ 𝑛 ≤ 10 and 𝑥 𝑛 = 𝑢 𝑛 − 𝑢 𝑛 − 4 . Plot
ℎ[𝑛], 𝑥[𝑛] and y n = 𝑥 𝑛 ∗ ℎ[𝑛] for 0 ≤ 𝑛 ≤ 20 in three
subplots.

信号与系统实验 Signals and Systems Laboratory 12

信号与系统实验 Signals and Systems Laboratory 13
Part 2. Causal LTI Systems
Described by Linear Constant-
Coefficient Difference Equations

信号与系统实验 Signals and Systems Laboratory 14

 Linear Constant-Coefficient Difference Equations
𝑁 𝑀

𝑎𝑘 𝑦 𝑛 − 𝑘 = 𝑏𝑘 𝑥[𝑛 − 𝑘]
𝑘=0 𝑘=0

In order to completely define the input-output relationship of a system, some

auxiliary conditions are required. Different choices of the auxiliary conditions lead
to different input-output relationships. For an LTI system, the causality condition is
equivalent to the condition of initial rest—i.e., if 𝑥 𝑛 = 0 for 𝑛 < 𝑛0 , then 𝑦 𝑛 =
0 for all 𝑛 < 𝑛0 as well.

Linear constant-
coefficient difference + The condition
= Causal LTI
of initial rest system
equation

信号与系统实验 Signals and Systems Laboratory 15

 Linear Constant-Coefficient Difference Equations
𝑁 𝑀

𝑎𝑘 𝑦 𝑛 − 𝑘 = 𝑏𝑘 𝑥[𝑛 − 𝑘]
𝑘=0 𝑘=0
𝑀
𝑏𝑘  Non-recursive
𝑁=0 𝑦𝑛 =
𝑎0
𝑥[𝑛 − 𝑘]  No need for auxiliary
𝑘=0
conditions
 Impulse response has
FIR 𝑏𝑛 finite duration
0≤𝑛≤𝑀
ℎ 𝑛 = 𝑎0  The output can be
calculated by conv(x,h)
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
Example: 𝑦 𝑛 = 𝑥 𝑛 + 0.5𝑥 𝑛 − 1 + 2𝑥[𝑛 − 2]
 Recursive
 Auxiliary conditions is
𝑁≥1 Impulse response has
required
infinite duration  Impulse response has
IIR infinite duration
Example: 𝑦 𝑛 − 0.8𝑦[𝑛 − 1] = 2𝑥 𝑛  The output cannot be
calculated by conv(x,h)

信号与系统实验 Signals and Systems Laboratory 16

 Causal LTI Systems
A causal discrete-time LTI system is uniquely determined by two vectors
𝐴 = [𝑎0 𝑎1 𝑎2 ⋯ 𝑎𝑁 ] and 𝐵 = [𝑏0 𝑏1 ⋯ 𝑏𝑀 ].

𝐴 = [1], 𝐵 = [1 0.5 2] for 𝑦 𝑛 = 𝑥 𝑛 + 0.5𝑥 𝑛 − 1 + 2𝑥[𝑛 − 2]

𝐴 = [1 − 0.8], 𝐵 = [2] for 𝑦 𝑛 − 0.8𝑦[𝑛 − 1] = 2𝑥 𝑛

How to define the system?

𝑦 𝑛 = −𝑦 𝑛 − 1 − 0.5𝑦 𝑛 − 2 + 0.2𝑥 𝑛 + 0.1𝑥 𝑛 − 1 + 0.1𝑥[𝑛 − 2]

信号与系统实验 Signals and Systems Laboratory 17

 Output Calculation for Discrete-Time
Causal LTI Systems
For FIR systems, the command 𝑦 = 𝑐𝑜𝑛𝑣(𝑥, ℎ) can be
utilized, but how about the IIR system?

𝑦 = 𝑓𝑖𝑙𝑡𝑒𝑟(𝐵, 𝐴, 𝑥)

This is a solution for all discrete-time causal LTI

systems including FIR and IIR systems.

Attention: y share the same time index with the input

x, so the output y may be truncated, it does not cover
the whole nonzero elements of the output.

信号与系统实验 Signals and Systems Laboratory 18

 Output Calculation for Discrete-Time
Causal LTI Systems
𝑦 𝑛 − 0.8𝑦[𝑛 − 1] = 2𝑥 𝑛

A = [1 -0.8];
B = 2;
x1 = [1]; % input is unit impulse,
output is the unit impulse response
y1 = filter(B,A,x1); y1n = 0;
x2 = [1 0 0 0];
y2 = filter(B,A,x2); y2n = 0:3;
x3 = [1 0 0 0 0 0 0 0 0 0 0];
y3 = filter(B,A,x3); y3n = 0:10;
subplot(3,1,1); stem(y1n,y1); xlim([0,10]);
subplot(3,1,2); stem(y2n,y2); xlim([0,10]);
subplot(3,1,3); stem(y3n,y3); xlim([0,10]);

MATLAB assume 𝑦 −1 = 0

信号与系统实验 Signals and Systems Laboratory 19

 Output Calculation for Discrete-Time
Causal LTI Systems
𝑦 𝑛 = 𝑥 𝑛 + 0.5𝑥 𝑛 − 1 + 2𝑥[𝑛 − 2]

A = 1;B = [1 0.5 2]; % h[n] = B[n],

nh = 0:2, Nh = 3
x = [1]; % nx = 0:0
xx = [1 0 0]; % extend x[n], add
(Nh-1) zeros in the behind.
y1 = filter(B,A,xx); y1n = 0:2;
y2 = conv(x,B); y2n = 0:2;
subplot(2,1,1); stem(y1n,y1);
xticks([0:5]); xlim([0,5]);
title('impulse response by
filter');
subplot(2,1,2); stem(y2n,y2,'r');
xticks([0:5]); xlim([0,5]);
title('impulse response by conv');
MATLAB assume 𝑥 −1 = 𝑥[−2] = 0

信号与系统实验 Signals and Systems Laboratory 20

 Output Calculation for Discrete-Time
Causal LTI Systems
Steps for calculate the output of FIR systems
𝑀

yn = 𝑏𝑚 𝑥[𝑛 − 𝑚]
𝑚=0

① Define ℎ = 𝑏0 𝑏1 ⋯ 𝑏𝑀 and its time index 𝑛ℎ = 0: 𝑀;

② Define input sequence 𝑥[𝑛] and its time index 𝑛𝑥;
③ Extend the input sequence by 𝑀 zeros 𝑥 = 𝑥 𝑧𝑒𝑟𝑜𝑠 1, 𝑀 , update
its time index simultaneously 𝑛𝑥 = 𝑛𝑥 1 : 𝑛𝑥 𝑒𝑛𝑑 + 𝑀
④ Calculate output sequence by 𝑦 = 𝑓𝑖𝑙𝑡𝑒𝑟 ℎ, 1, 𝑥 ;
⑤ Define the time index of the output by 𝑛𝑦 = 𝑛𝑥;
⑥ Plot the output of the system by 𝑠𝑡𝑒𝑚(𝑛𝑦, 𝑦).

信号与系统实验 Signals and Systems Laboratory 21

 Comparison Between conv(u,v) and
filter(b,a,x)

𝒘 = 𝒄𝒐𝒏𝒗(𝒖, 𝒗) 𝒚 = 𝒇𝒊𝒍𝒕𝒆𝒓(𝒃, 𝒂, 𝒙)
Signal of finite duration Signal of finite duration
FIR FIR and IIR
𝑛𝑤 𝑛𝑦 = 𝑛𝑥
= 𝑛𝑢 1 Truncated output
+ 𝑛𝑣 1 : 𝑛𝑢 𝑒𝑛𝑑
+ 𝑛𝑣(𝑒𝑛𝑑)

信号与系统实验 Signals and Systems Laboratory 22

 Impulse Response for Discrete-Time Causal
LTI Systems by impz(b,a,k)

𝑦 𝑛 + 0.5𝑦 𝑛 − 1 − 0.2𝑦[𝑛 − 2] = 10𝑥 𝑛

a = [1 0.5 -0.2];
b = 10;
k = 20;
impz(b,a,k)

Number
of points
MATLAB assume 𝑦 −2 = 𝑦[−1] = 0

信号与系统实验 Signals and Systems Laboratory 23

 Exercise 2.2
𝑦 𝑛 = −𝑦 𝑛 − 1 − 0.5𝑦 𝑛 − 2 + 0.2𝑥 𝑛 + 0.1𝑥 𝑛 − 1 + 0.1𝑥[𝑛 − 2]

For the above system, compute and plot in the time interval
0 ≤ 𝑛 ≤ 20

a) The impulse response of the system

b) The step response of the system
c) The response of the system to the input 𝑥 𝑛 = 𝑛 ∙ 0.9𝑛 ,
0≤𝑛≤5

信号与系统实验 Signals and Systems Laboratory 24

信号与系统实验 Signals and Systems Laboratory 25
Part 3. Convolution Integral for
Continuous-Time LTI Systems

信号与系统实验 Signals and Systems Laboratory 26

 Representation of Continuous-Time Signals

+∞

𝑥(𝑡) = 𝑥(𝑘∆)𝛿∆ (𝑡 − 𝑘∆)∆

𝑘=−∞

𝛿 𝑡 = lim 𝛿∆ (𝑡)
∆→0 𝑥(𝑘∆)𝛿∆ (𝑡 − 𝑘∆)∆

+∞
𝑥 𝑡 = lim 𝑥 𝑡 = 𝑥 𝜏 𝛿 𝑡 − 𝜏 𝑑𝜏
∆→0 −∞

信号与系统实验 Signals and Systems Laboratory 27

 Continuous-Time Unit Impulse Response

𝛿∆ (𝑡 − 𝑘∆) S ℎ𝑘∆ (𝑡) (𝑘 = ⋯ , −1,0,1, ⋯)

S is linear
+∞ +∞

𝑥(𝑡) = 𝑥(𝑘∆)𝛿∆ (𝑡 − 𝑘∆)∆ S 𝑦(𝑡) = 𝑥 𝑘∆ ℎ𝑘∆ (𝑡)∆

𝑘=−∞ 𝑘=−∞
+∞ +∞
𝑥 𝑡 = lim 𝑥 𝑡 = 𝑥 𝜏 𝛿 𝑡 − 𝜏 𝑑𝜏 𝑦 𝑡 = lim 𝑦 𝑡 = 𝑥 𝜏 ℎ𝜏 𝑡 𝑑𝜏
∆→0 ∆→0 −∞
−∞
S is time-invariant
ℎ𝜏 (𝑡) = ℎ(𝑡 − 𝜏)

+∞ +∞
𝑥 𝑡 = 𝑥 𝜏 𝛿 𝑡 − 𝜏 𝑑𝜏 S 𝑦 𝑡 = 𝑥 𝜏 ℎ 𝑡 − 𝜏 𝑑𝜏
−∞ −∞

Convolution Integral!

信号与系统实验 Signals and Systems Laboratory 28

Response of LTI Systems: Convolution Integral
+∞
𝑦 𝑡 = 𝑥 𝜏 ℎ 𝑡 − 𝜏 𝑑𝜏
−∞

𝑦(𝑡) = 𝑥(𝑡) ∗ ℎ(𝑡)

1. An LTI system is completely characterized by

its impulse response h(t).
2. From the impulse response, we can know
everything about the LTI system. (causality,
stability, static/dynamic, invertibility)
3. For any input, we can find the output
immediately by the convolution integral.

信号与系统实验 Signals and Systems Laboratory 29

 Calculate Convolution Integral
Example 𝑦(𝑡) = 𝑥(𝑡) ∗ ℎ(𝑡) 1

h(t) = 𝑒 −𝑡 𝑢(𝑡) 𝑥(𝑡) = 𝑢(𝑡) 0 

1
1 1 1 
0
  t
0 0 0 1

① Flip ℎ(𝜏) to ℎ(−𝜏); 0
② Shift ℎ(−𝜏) to ℎ(𝑡 − 𝜏) for any t;
③ Multiply 𝑥(𝜏) with ℎ(𝑡 − 𝜏) over the whole 𝜏 axis; 1
−(𝑡−𝜏) 𝑡 > 0
𝑥(𝜏)ℎ(𝑡 − 𝜏) = 𝑒 
0 𝑡≤0 t 0
⑤ Integrate over 𝜏, obtain 𝑦(𝑡)
1

0 t

信号与系统实验 Signals and Systems Laboratory 30

 Convolution Integral in MATLAB: conv(u,v)
+∞
y 𝑡 = −∞
𝑢 𝜏 𝑣 𝑡 − 𝜏 𝑑𝜏
1−𝑡 0≤𝑡≤1
𝑢 𝑡 =
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
y(𝑡) = 𝑢(𝑡) ∗ 𝑣(𝑡)
1 0≤𝑡≤2
 𝑣 𝑡 =
Define time step length: step_length 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
 Define 𝑢 𝑡𝑢 and v, 𝑡𝑣;
 Calculate y(𝑡) by y = 𝑐𝑜𝑛𝑣(𝑢, 𝑣)*step_length;
 Calculate the time index for nonzero values of y(𝑡)
𝑛𝑦 = 𝑛𝑢 1 + 𝑛𝑣 1 : 𝑛𝑢 𝑒𝑛𝑑 + 𝑛𝑣(𝑒𝑛𝑑)
step_length = 0.01;
tv = 0: step_length :2;
v = ones(size(tv));
tu = 0: step_length :1;
u = 1-tu;
y = conv(u,v)* step_length;
ty = 0: step_length :3;
figure; plot(ty,y);

信号与系统实验 Signals and Systems Laboratory 31

 Exercise 2.3
Calculate the convolution integral between the continuous-
time signals ℎ 𝑡 = 𝑒 −𝑡 𝑢 𝑡 − 𝑢(𝑡 − 10) and
0.6 −1 ≤ 𝑡 < 0.5
𝑥 𝑡 = 0.3 0.5 ≤ 𝑡 < 3
0 𝑡 < −1 𝑎𝑛𝑑 𝑡 > 3
Plot ℎ(𝑡), 𝑥(𝑡) and y(t) = 𝑥(𝑡) ∗ ℎ(𝑡) for −1 ≤ 𝑡 ≤ 20 in
three subplots.

信号与系统实验 Signals and Systems Laboratory 32

Part 4. Causal LTI Systems
Described by Linear Constant-
Coefficient Differential Equations

信号与系统实验 Signals and Systems Laboratory 33

 Linear Constant-Coefficient Differential Equations
𝑁 𝑀
𝑑𝑘 𝑦(𝑡) 𝑑 𝑘 𝑥(𝑡)
𝑎𝑘 = 𝑏𝑘
𝑑𝑡𝑘 𝑑𝑡𝑘
𝑘=0 𝑘=0

In order to completely define the input-output relationship of a system, some

auxiliary conditions are required. Different choices of the auxiliary conditions lead
to different input-output relationships. For an LTI system, the causality condition is
equivalent to the condition of initial rest—i.e., if 𝑥(𝑡) = 0 for t ≤ 𝑡0 , then 𝑦(𝑡) = 0
for all t ≤ 𝑡0 as well, and therefore, the response for 𝑡 > 𝑡0 can be calculated from
the differential equation with the initial conditions
𝑑𝑦(𝑡0 ) 𝑑 𝑁−1 𝑦 𝑡0
𝑦 𝑡0 = =⋯= =0
𝑑𝑡 𝑑𝑡𝑁−1

Linear constant-
coefficient differential + The condition
= Causal LTI
of initial rest system
equation

信号与系统实验 Signals and Systems Laboratory 34

Linear Constant-Coefficient Differential Equations
𝑁 𝑀
𝑑 𝑘 𝑦(𝑡) 𝑑 𝑘 𝑥(𝑡)
𝑎𝑘 = 𝑏𝑘
𝑑𝑡𝑘 𝑑𝑡𝑘
𝑘=0 𝑘=0

𝑀
1 𝑑 𝑘 𝑥(𝑡)  Differentiate the input
𝑁=0 𝑦(𝑡) =
𝑎0
𝑏𝑘
𝑑𝑡𝑘
explicitly
𝑘=0
 No need for auxiliary
conditions

Example: 𝑦(𝑡) = 𝑥(𝑡) + 0.5𝑥(𝑡) + 2𝑥(𝑡)

 Implicit
 Auxiliary conditions is
𝑁≥1 𝑦 𝑡 = 𝑦ℎ 𝑡 + 𝑦𝑝 (𝑡) required
 The response consists
of two parts— a
Example: 𝑦(𝑡) − 0.8𝑦(𝑡) = 2𝑥(𝑡) particular solution plus
a homogeneous
solution

信号与系统实验 Signals and Systems Laboratory 35

 Causal LTI Systems
A causal continuous-time LTI system is uniquely determined by two
vectors 𝐴 = [𝑎𝑁 ⋯ 𝑎2 𝑎1 𝑎0 ] and 𝐵 = [𝑏𝑀 ⋯ 𝑏1 𝑏0 ].

𝐴 = [1], 𝐵 = [2 0.5 1] for 𝑦(𝑡) = 𝑥(𝑡) + 0.5𝑥(𝑡) + 2𝑥(𝑡)

𝐴 = [1 − 0.8], 𝐵 = [2] for 𝑦(𝑡) − 0.8𝑦(𝑡) = 2𝑥(𝑡)

How to define the system?

𝑦 𝑡 + 𝑦 𝑡 + 0.5𝑦 𝑡 = 0.2𝑥 𝑡 + 0.1𝑥 𝑡 + 0.1𝑥(𝑡)

Pay attention to the order of the

coefficient vectors here!

信号与系统实验 Signals and Systems Laboratory 36

 Output Calculation for Continuous-Time
Causal LTI Systems

𝑦 = 𝑙𝑠𝑖𝑚 𝐵, 𝐴, 𝑥, 𝑡
𝑦 = 𝑙𝑠𝑖𝑚(𝑠𝑦𝑠, 𝑥, 𝑡) Transfer
function
𝑠𝑦𝑠 = 𝑡𝑓(𝐵, 𝐴)
Impulse response: 𝑦 = 𝑖𝑚𝑝𝑢𝑙𝑠𝑒 𝐵, 𝐴, 𝑡
Step response: 𝑦 = 𝑠𝑡𝑒𝑝(𝐵, 𝐴, 𝑡)

𝐴 = [𝑎𝑁 ⋯ 𝑎2 𝑎1 𝑎0 ] 𝐵 = [𝑏𝑀 ⋯ 𝑏1 𝑏0 ]

𝑁 𝑀
𝑑 𝑘 𝑦(𝑡) 𝑑 𝑘 𝑥(𝑡)
𝑎𝑘 = 𝑏𝑘
𝑑𝑡𝑘 𝑑𝑡𝑘
𝑘=0 𝑘=0

信号与系统实验 Signals and Systems Laboratory 37

 Output Calculation for Continuous-Time
Causal LTI Systems
𝑦 𝑡 + 2𝑦 𝑡 + 10𝑦 𝑡 = 10𝑥(𝑡)
a = [1 2 10]; b = 10;
t = 0:0.01:10; x = cos(2*pi*t);
y = lsim(b,a,x,t);
figure(1);impulse(b,a,t); MATLAB assume 𝑦(0) = 𝑦(0) = 0
figure(2);step(b,a,t);
figure(3);plot(t,y); title('response to the input
x(t)=cos(2\pit)');

信号与系统实验 Signals and Systems Laboratory 38

 Exercise 2.4
For the system with the condition of initial rest,
𝑦 𝑡 + 3𝑦 𝑡 + 12𝑦 𝑡 = 𝑥 𝑡 − 𝑥 𝑡 − 6𝑥(𝑡)
Calculate and plot the impulse response, step response
and the response to the input
𝑥 𝑡 = cos 3𝜋𝑡 𝑢(𝑡)
for 0 ≤ 𝑡 ≤ 10.

信号与系统实验 Signals and Systems Laboratory 39

信号与系统实验 Signals and Systems Laboratory 40
Part 5. Design a Wireless Signal
Detector

信号与系统实验 Signals and Systems Laboratory 41

 Example: Simplified Wireless Channel
Attenuation 0.4; delay 3

Attenuation 0.8; delay 1

Transmitter Receiver
Attenuation 0.6; delay 2
Wireless Channel (LTI System)
𝑦 𝑛 = 0.8𝑥 𝑛 − 1 + 0.6𝑥 𝑛 − 2 + 0.4𝑥[𝑛 − 3]

Detected Signal Received Signal

Signal Detector
z 𝑛 = 𝑥[𝑛] or 𝑦[𝑛]
(LTI System)
𝑧 𝑛 = 𝑥[𝑛 − 𝑘]
Impulse Response: ℎ2 𝑛 ∗ ℎ1 𝑛 = 𝛿[𝑛] or 𝛿 𝑛 − 𝑘
Difference Equation: 0.8𝑧 𝑛 − 1 + 0.6𝑧 𝑛 − 2 + 0.4𝑧 𝑛 − 3 = 𝑦[𝑛]
𝑧 𝑛 = 𝑧[𝑛 − 1] → 0.8𝑧 𝑛 + 0.6𝑧 𝑛 − 1 + 0.4𝑧 𝑛 − 2 = 𝑦[𝑛]

信号与系统实验 Signals and Systems Laboratory 42

 Wireless Channel Simulation
Attenuation 0.4; delay 3

Attenuation 0.8; delay 1

Transmitter Receiver
Attenuation 0.6; delay 2
Wireless Channel (LTI System)

% Generate Tx signal
x = [1 2 3 4 0 0 0 1 2 5 0 0 ];
% Obtian Rx signal by simulating the wireless channel
a1 = 1;
b1 = [0 0.8 0.6 0.4];
y = filter(b1,a1,x);
% Compare the transmitted signal with the received
signal
figure(1);
subplot(2,1,1); stem(x); title('transmitted signal
x[n]');
subplot(2,1,2); stem(y); title('received signal
y[n]');

信号与系统实验 Signals and Systems Laboratory 43

 Signal Detector Design Verification
Detected Signal Received Signal
Signal Detector
or
(LTI System)

Impulse Response: or
Difference Equation:

% Generate the detected signal by

the design of a detector
a2 = [0.8 0.6 0.4]; b2 = 1;
z = filter(b2,a2,y);
% Compare the transmitted signal
with the detected signal
figure(2);
subplot(2,1,1); stem(x);
title('transmitted signal x[n]');
subplot(2,1,2); stem(z);
title('detected signal z[n]');

信号与系统实验 Signals and Systems Laboratory 44

 Think More…
In the above example, the wireless channel model is known,
the filter structure and parameters are given. The detector
design problem is a piece of cake, just use inverse filtering, find
the inverse of the impulse response of the original wireless
channel model.
If the wireless channel model is unknown, the problem is
very difficult, but if we know the structure of the model, but the
parameters are to be determined, then this is a relatively easy
problem: parameter estimation problem.
How can we find the parameters in the model?
𝑦 𝑛 = 𝑥 𝑛 + 𝛼𝑥[𝑛 − 𝑁]
How to find 𝛼? Least squares? How to find N? That is exactly
what the assignment 2.10 needs you to do! Do we need any
tools? Any mathematics?

信号与系统实验 Signals and Systems Laboratory 45

 Echo Cancellation (Lab Assignment 2.10)

𝑅𝑦𝑦 𝑛 = 𝑥[𝑛] ∗ 𝛿 𝑛 + 𝛼𝛿 𝑛 − 𝑁 ∗ 𝑥[−𝑛] ∗ 𝛿 𝑛 + 𝛼𝛿 𝑛 + 𝑁

= 𝑅𝑥𝑥 [𝑛] ∗ 1 + 𝛼 2 𝛿 𝑛 + 𝛼𝛿 𝑛 − 𝑁 + 𝛼𝛿[𝑛 + 𝑁]
= 1 + 𝛼 2 𝑅𝑥𝑥 𝑛 + 𝛼𝑅𝑥𝑥 𝑛 − 𝑁 + 𝛼𝑅𝑥𝑥 [𝑛 + 𝑁]

𝑅𝑥𝑦 𝑛 = 𝑥 𝑛 ∗ 𝑦 −𝑛 = 𝑥[𝑛] ∗ 𝑥[−𝑛] ∗ 𝛿 𝑛 + 𝛼𝛿 𝑛 + 𝑁

= 𝑅𝑥𝑥 [𝑛] ∗ 𝛿 𝑛 + 𝛼𝛿[𝑛 + 𝑁]
= 𝑅𝑥𝑥 𝑛 + 𝛼𝑅𝑥𝑥 [𝑛 + 𝑁]

信号与系统实验 Signals and Systems Laboratory 46

 Echo Cancellation (Lab Assignment 2.10)

Attenuation 𝛼; delay N

Attenuation 1; delay 0
Transmitter Receiver
𝑦 𝑛 = 𝑥 𝑛 + 𝛼𝑥 𝑛 − 𝑁

𝑦 𝑛 = 𝑥 𝑛 + 𝛼𝑥 𝑛 − 𝑁

min 𝑦 𝑛 − 𝑦[𝑛]
𝛼,𝑁

信号与系统实验 Signals and Systems Laboratory 47

 Auto-correlation and Cross-correlation
Auto-correlation: 𝜙𝑥𝑥 𝑛 = +∞
𝑚=−∞ 𝑥 𝑚 + 𝑛 𝑥 𝑚 = 𝑥[𝑛] ∗ 𝑥[−𝑛]

Cross-correlation: 𝜙𝑥𝑦 𝑛 = +∞
𝑚=−∞ 𝑥 𝑚 + 𝑛 𝑦 𝑚 = 𝑥[𝑛] ∗ 𝑦[−𝑛]

clear all; close all;

u = randn(1,40); nu = 1:40;
v = fliplr(u); nv = -40:-1;
w = conv(u,v);
nw = nu(1)+nv(1):nu(end)+nv(end);
figure(1);
subplot(2,1,1); stem(nu,u);
xlabel('n');ylabel('u[n]');
title('a random signal');
subplot(2,1,2);stem(nw,w);
xlabel('n');ylabel('\phi_{uu}');
title('auto-correlation of a
random signal');

Auto-correlation of a random signal has a high peak at the origin.

信号与系统实验 Signals and Systems Laboratory 48

 Auto-correlation Application: Echo Cancellation
Auto-correlation: 𝜙𝑥𝑥 𝑛 = +∞
𝑚=−∞ 𝑥 𝑚 + 𝑛 𝑥 𝑚 = 𝑥[𝑛] ∗ 𝑥[−𝑛]

Cross-correlation: 𝜙𝑥𝑦 𝑛 = +∞
𝑚=−∞ 𝑥 𝑚 + 𝑛 𝑦 𝑚 = 𝑥[𝑛] ∗ 𝑦[−𝑛]
nx = 100; x = randn(1,nx);
N = 50; alpha = 0.9; 𝑦 𝑛 = 𝑥 𝑛 + 0.9𝑥 𝑛 − 50
y = filter([1 zeros(1,N-1) alpha],1,x);
Rxx = conv(x,fliplr(x));
Ryy = conv(y,fliplr(y));
Ryx = conv(y,fliplr(x));
figure;
subplot(3,1,1); plot([-nx+1:nx-1],Rxx);grid on;
title('autocorrelation of x[n]');
subplot(3,1,2); plot([-nx+1:nx-1],Ryy);grid on;
title('autocorrelation of y[n]');
subplot(3,1,3); plot([-nx+1:nx-1],Ryx);grid on;
title('cross-correlation of y[n] and x[n]');

Look at the differences between auto-correlations and cross-correlation

of input and output signals of a filter.

信号与系统实验 Signals and Systems Laboratory 49

 Lab#2 Assignment
 Read tutorial 2.1&2.2&2.3 by yourself
 Finish 2.4 & 2.5 & 2.10
◇ The sound file for 2.10: lineup.mat
Download it from Blackboard.

Deadline: 10:00am October 14th

Submit the report in word or pdf
format through Blackboard.

信号与系统实验 Signals and Systems Laboratory 50